Final Activity Report Summary - QMC-DFT (Development and application of electronic quantum Monte Carlo methods to improve density functional theory.)
The project concerned developments in computational electronic structure theory which were relevant to theoretical chemistry and materials science. It aimed to allow the fellow to pursue the work he began towards the improvement of the multi-configurational density functional theory (DFT) method that was proposed by Dr Andreas Savin and co-workers. This theory had the potential of curing the main shortcomings of present-day DFT by combining a long-range conventional wave function calculation with a short-range density functional approximation.
More specifically, the goal of the project was to extend to general many-electron systems the recently proposed Overhauser method which, starting from approximate effective two-electron potentials, could produce very accurate short-range system averaged pair densities, also called intracule densities, and therefore very accurate short-range correlation energies, which were needed in the multi-configurational DFT approach. For this purpose, the adopted strategy was to develop and apply appropriate quantum Monte Carlo (QMC) techniques in order to generate the accurate reference data relevant to the Overhauser method.
As planned, taking advantage of the expertise in QMC techniques at the outgoing institution, Cornell University of the United States of America, we developed a robust and efficient method to optimise all parameters in large multideterminant QMC wave functions based on energy minimisation. We demonstrated that this method made it possible to obtain molecular dissociation energies with near chemical accuracy. The work had an important impact in QMC community and several research groups in the world had already started using our optimisation method by the time of the project completion.
As also planned, we developed improved statistical estimators which permitted calculations of intracule densities that were several orders of magnitude more efficient than the usual Monte Carlo approach used for this kind of calculations. Thanks to these improved estimators, along with the achievement of systematically reducing the systematic error due to wave function by optimisation of an increasing number of parameters, we obtained accurate correlated intracule densities for atoms and molecules.
Finally, the fellow was hired as a faculty member at the return institution, Pierre and Marie Curie University located in Paris, France, and he was already pursuing the work on using these obtained accurate reference data to extend the Overhauser method.
More specifically, the goal of the project was to extend to general many-electron systems the recently proposed Overhauser method which, starting from approximate effective two-electron potentials, could produce very accurate short-range system averaged pair densities, also called intracule densities, and therefore very accurate short-range correlation energies, which were needed in the multi-configurational DFT approach. For this purpose, the adopted strategy was to develop and apply appropriate quantum Monte Carlo (QMC) techniques in order to generate the accurate reference data relevant to the Overhauser method.
As planned, taking advantage of the expertise in QMC techniques at the outgoing institution, Cornell University of the United States of America, we developed a robust and efficient method to optimise all parameters in large multideterminant QMC wave functions based on energy minimisation. We demonstrated that this method made it possible to obtain molecular dissociation energies with near chemical accuracy. The work had an important impact in QMC community and several research groups in the world had already started using our optimisation method by the time of the project completion.
As also planned, we developed improved statistical estimators which permitted calculations of intracule densities that were several orders of magnitude more efficient than the usual Monte Carlo approach used for this kind of calculations. Thanks to these improved estimators, along with the achievement of systematically reducing the systematic error due to wave function by optimisation of an increasing number of parameters, we obtained accurate correlated intracule densities for atoms and molecules.
Finally, the fellow was hired as a faculty member at the return institution, Pierre and Marie Curie University located in Paris, France, and he was already pursuing the work on using these obtained accurate reference data to extend the Overhauser method.